How to query data using natural language - intro to AI features in Oracle 23ai
summary
Demo shows Oracle 23ai running Select AI on an Autonomous Database hooked to OCI Generative AI, generating SQL from natural language and exposing the actual SQL with show SQL and explain SQL. It embeds text with a 384‑dim model inside the DB, indexes ~3,000 beers for vector search, and runs similarity queries; conversational mode retains up to 10 prompts. Gotchas include hallucinated gender inference, wrong min/max department ID results, and a measured cost of $1.86 for hundreds of queries.
AI-written summary · about
chapters (11)
- 0:00Introduction and CERN context
- 1:32Getting started with 23ai
- 3:04Understanding select AI
- 5:22Setting up select AI
- 8:17Select AI demo and analysis
- 14:03Advanced querying and challenges
- 25:32Flexibility and cost analysis
- 28:28Vector theory for AI
- 34:08Beer recommendation system demo
- 38:17Real-world RAG applications
- 44:44Closing and Q&A
related talks
12mTackle Any Workload with Oracle Database, AI, and Multicloud
23mBuild Natural Language-Powered Apps with Select AI
17mBuild GenAI Apps with Oracle Database 23ai
23mAI for Data Made Simple
12mKey AI Innovations in Oracle Database 23ai
48mOracle Database Directions: Innovate, Scale, and Sleep Easier
01h 21mKeynote on the "AI for Data" Revolution
18m3 Oracle AI Use Cases